You’ve heard of the raw foods movement? Well, we here at SurveyMonkey are officially launching the raw data movement. We’re pleased to announce that our Internet poll successfully predicted the winners of the gubernatorial elections in Virginia and New Jersey!
So we got the winners right, but how did we do on the actual percentages?
To get our specific predictions, we just added all of our ‘still undecided’ but ‘definitely voting’ respondents to the total for the candidate party that traditionally wins in that state. So, in Virginia we added them to the Republican candidate’s tally and in New Jersey we added them to the Democratic candidate’s tally.
As you can see, we did pretty well for ourselves. We were within one percentage point for New Jersey and within three percentage points for Virginia.
The place where we had the most trouble was in overestimating the percentage of people who would be voting for the Libertarian candidate–Robert Sarvis–in Virginia. This is an interesting issue because polling numbers were used to determine who would be a legitimate participant in Virginia’s third gubernatorial debate. Sarvis was ultimately excluded from this debate because he failed to meet a 10% polling threshold in RealClearPolitics (RCP) polls. At the time, RCP had Sarvis polling at 9.9%, while we showed him polling at 10.01%. Ultimately, our results show, however, that this projected support was inflated.
This phenomenon is fairly common in political contests. Although independent candidates sound good in theory, when a voter gets to the polls, it can be tougher to cast that vote. The problem is, the independent candidate almost always loses, so one’s vote becomes a statement, a gesture–there’s frequently almost a feeling that it doesn’t quite count the same way that a vote for a Democratic or Republican candidate would. Consequently, the actual voting data suggests that some of those independent voters didn’t actually end up getting to the polls at all (even though they said they would) or that when they got there, they voted for the Democratic candidate instead. Nevertheless, our raw estimates were just as accurate as RCP’s weighted estimates.
So here’s to another successful election cycle. We’re proud to say that both of our predicted winners both came through. And this time, we did it without any weighting. That’s right. We just predicted an election outcome with raw data. Try doing that with a phone poll!
Questions for Liana? Leave ’em for her below!